| This dissertation (a) hypothesizes that information flow plays an important role in the inter-temporal variation in stock return, and (b) describes a surprising absence of scale for forecast error and forecast dispersion distributions.In "Information Flow and Stock Returns", I propose an information flow explanation for the "Monday effect," defined as higher stock returns on Friday than on Monday. I hypothesize that short sellers profit when negative information supporting their bearish outlook is released to the public. Such public information is more likely to arrive on days with higher information flow (e.g., when the stock market is open). The cost of short selling, however, is a function of the loan rate which is invariant to whether markets are closed or open. This creates an incentive for short sellers to close their short positions on Fridays, relative to other days. Closing a short position requires the short seller to place a buy order. As a result, the buying pressure on Friday raises share prices slightly.Turning to my second paper, "Surprising absence of scale for forecast error and forecast dispersion distributions" (with Jake Thomas), we show that while levels of actual and consensus forecast earnings per share (EPS) vary with scale, magnitudes of the difference (or forecast errors) do not vary with scale. That is, forecast errors within a certain range (e.g., +/-5 cents per share) are equally likely for both high-price and low-price shares.We also find a similar lack of variation with scale for forecast dispersion, representing magnitudes of the difference between individual forecasts and the consensus (mean) for that firm-quarter. The prior literature has assumed that magnitudes of forecast errors (representing predictability) and forecast dispersion (representing disagreement across analysts) vary naturally with scale and has deflated both variables accordingly. We show that such scaling is likely to cause biased estimates, and recommend that scaling not be used unless called for by theory, and a scale variable be included as an additional regressor. Our exploratory analyses suggest that both variables vary with scale but other effects that are correlated with scale reverse that variation. |
